{"id":"W4385573178","doi":"10.18653/v1/2022.blackboxnlp-1.10","title":"How (Un)Faithful is Attention?","year":2022,"lang":"en","type":"article","venue":"","topic":"Explainable Artificial Intelligence (XAI)","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Concordia University","funders":"","keywords":"Computer science; Metric (unit); Artificial intelligence; Binary number; Encoder; Simple (philosophy); Measure (data warehouse); Machine learning; Data mining; Mathematics; Arithmetic","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.0002920681,0.00009423265,0.00008984752,0.0001053025,0.0005240138,0.0003429409,0.001196167,0.00001815163,0.00139537],"category_scores_gemma":[0.00001934218,0.00009449626,0.00007731962,0.0006985979,0.00002647371,0.0006917381,0.0008332622,0.0001433439,0.0003612113],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00006492569,"about_ca_system_score_gemma":0.00003794115,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00007188321,"about_ca_topic_score_gemma":0.00001006609,"domain_scores_codex":[0.9986834,0.00007450187,0.0001306864,0.0003877092,0.0004214403,0.0003022546],"domain_scores_gemma":[0.9991429,0.00004056804,0.00005231697,0.0006226567,0.00006618861,0.00007539654],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"not_applicable","study_design_scores_codex":[0.000003439101,0.0001594351,0.001065888,0.000004901658,0.00002241133,0.00007112453,0.001542211,0.0001019709,0.00333846,0.8581803,0.04027032,0.09523955],"study_design_scores_gemma":[0.0002583683,0.0004733501,0.001736609,0.000004686774,0.00001151075,0.0001345326,0.006268966,0.2097873,0.08223373,0.1303595,0.5678537,0.0008777318],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.03881933,0.00009702373,0.8826005,0.05355895,0.0008584239,0.0002241497,0.000003812559,0.000482231,0.02335556],"genre_scores_gemma":[0.9472011,0.000003590753,0.01609919,0.003178797,0.00004649827,0.00005935313,0.000001660378,0.000008803194,0.03340097],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9083818,"threshold_uncertainty_score":0.9995175,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02181393702308317,"score_gpt":0.2358932240045686,"score_spread":0.2140792869814855,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}